Microsoft announces Microsoft-Decision-1, a fast decision model it is testing internally

Microsoft has a decision model of its own. Satya Nadella announced Microsoft-Decision-1 on X at 18:37 UTC today, calling it Microsoft's "new model for fast decision-making". It is in internal testing: there is no public API, model id, price or published benchmark yet.

WHAT MICROSOFT SAYS

From Nadella's post: it "delivers top performance on structured decision tasks, outperforming both LLMs and other decision models in latency and quality", and Microsoft is testing it "for everything from incident response and quality control to scientific discovery." Those are Microsoft's own claims. The post names no tasks, scores or latency figures, and does not say which decision models it was measured against.

WHAT A DECISION MODEL IS

It does not write prose. It reads an input and returns a typed answer: the probability that something is true, one option from a fixed list, or a score. That is the shape routing, classification, moderation and automated checks need, and it is why these models are pitched on speed and cost per call rather than on chat quality.

WHO ELSE HAS ONE

Microsoft is the latest of several. TypeSafe AI's Jev has been on the Vercel AI Gateway since September 17 and Liquid AI's d1 since September 30, both by our own catalog reads. OpenAI opened its Decisions API in public beta on October 6, running GPT-6 Luna at $0.10 per million input tokens with no output charge.

WHAT WE DO NOT KNOW

When or where it ships, and at what price. It is not in the OpenRouter, models.dev or Vercel AI Gateway catalogs we read at 19:30 UTC, and Nadella gave no date. Until Microsoft publishes numbers, the comparison with LLMs and other decision models is Microsoft's claim, not a finding.